�&���!�������$�$���������
������������������������� ������������������������
�������������������������������
!��"��#�$�%���'()*���% ++ ��,�����#+%�%��-+ '()*
��������&.���.��������������������)/0/��--���1-���-��2��1���3,��4#�5���/6)7'
�1#1-������
�������� ������� ����������� �������������������� ������������������������������������ !
"#���$#%������������&��'���������(��������������������� %����������������)��������������������
�����*+*��������%�����"""�,������� ����������������,������� �+���������������������������"""
,������� ��������������������������������%���������������������- ��.������� %����&��'�����/�����
�0�%�1���� �&��'�����%����1� ��0��.������� %����&��'���������2�����%�*23(�����&��'���������/���
,%�4�'����������������� %� 5&��'����6���� ������,%����/����� '������&��'����������������� ��� ��������
�������������)���� �������������/������ ������ �0���/����������������' �0 ���������)�2������
�'���� ���)�2���'������7������������������������������������������������� �����������������
���������+�������� ������)
8������0����������.����������9�����.)����)��* ������������'��)�������������������7%�������7����
����������%����0��:������������7� ������������������'��������� ������%���� ������8������%������'��
�����������)
���#����#����#����8�������8 ��2�4��������3��������-��#
���������������������4��������������
�&���!��"��#�$�%�������'()*
�1#1-��6//)
9������65�9(5�)5�97
�����
.-��#� ��� �:���-�2�� 8��#��14���8� 4���� -��� ��� �������-��#� ���3-5� �� -�� � ����� �-��,8�-��4
�������-��-�3���#��������%�����8��������3%������4��������-�4�1���-����3����������%�����18������#���8�2�8
������%������������8���4�� ���-��%�����1�8��-������2���#�5��-��,8�-��4��������-��-�3���������,�1��)0
%�����������������1�8��-��-���3%����� ��4�;� ��������������'0�%�������� ��4�,���������-��-��������-���#8�5
��������������3%����� ���-��%�4��-�������-�������88� ���������3-��-������#�����:%�����������8����5�����-
�����-1����4�������8���� �2��5��������#���4������4��-�2�������-�4���,8������--����3- ����3;-%���������:�4
������-��:%8����</�%��������������2�������������3%����� ���-��%�����-�����1��8��#��14���8�4�������2��
-����#� �2�4����� ����� ���3-� ���#��� -%������5� ,1�� 4��������5� ����-� ��� ��3%���� � ���-��%5� ���� ������-
4����3�������3-=��%��3�8����#���4�����->�!�����4������,���4��3��2�81���-�����3%�������4����3�����
�������-��-� ������#��,���4��3��2�81�5��-�3��-1��4�,���42����-��#����-����3�?���3�4����:%��4��1��-5
���#�����#������-������3%����� ���-��%��!����#1����������#����#���#������-������3%����� ���-��%��-
4�-���,8����������-� ����3����2�81�,8��,���4-�,���1-����4�2�41�8��������-��-���2���������2�-�����������4�
���,���4��3��2�81������-�@1���8�5���#�;2�81���������-��-����4�����:����3����4������3���#����8�������8
�2����1�8��-���������������������44�����5���#����3%����� ���-��%�����-�#�2���������-��-�,�������������2�-
���3������������2�81�����������,���4�
������������������ �������������
.��2��-�����������#���&1-���--������8 ����
A/)���%%��������� �����#����88���.��2��-���
������,��5���(')/< $���-,1�#�5�$��)06)7
��4��&�� ��4��&��
8��B13�����41 "������C-�� B������%�#�2
2
1. INTRODUCTION
In franchised chains, individual outlets are managed either by company managers or by
franchisees who own the outlet. When the outlet is operated corporately, the manager receives
mostly a salary, and internal mechanisms are used to monitor his behavior (such as personnel
evaluation and outlet-level performance data). When a franchisee owns a single outlet, he
typically acts as the general manager, and is the residual claimant on the outlet’s profits (minus
the payments to the franchisor). The franchisor in this case does not monitor directly the
franchisee’s behavior nearly as much as she does the company manager’s (see Lafontaine, 1992b
and Bradach, 1995 and 1997). Thus, within franchising firms, very different incentives and
monitoring mechanisms are used for people who are doing basically equivalent jobs.
Furthermore, franchising firms differ considerably in the degree to which they rely on company
ownership versus franchising — some chains are entirely franchised while others have a large
proportion of outlets owned and controlled by the company directly.
An extensive empirical literature has addressed already the reasons why franchising firms
choose the extent of company ownership that they do.1 However, prior studies have relied largely
on cross-sectional data. In this paper, we use an extensive panel data set we have developed to
address two main questions, namely: Do franchising firms change the extent to which they
control outlets directly over time as they learn more about the underlying factors that influence
the value of franchising? And secondly, why do some franchising firms choose to operate more
outlets directly than others?
1 Lafontaine (1992), Minkler and Park (1994), Scott (1995), and see Lafontaine and Slade (2001) for a review.
3
Our data set is a very unbalanced panel that follows almost 5000 U.S. and Canadian
franchisors for various intervals from 1980 through to 1997. Given this panel, we are able to do
two primary things that were not doable in past research. First, we are able to detail for the first
time the evolution of company ownership within firms, as they become older and better
established. Second, in assessing the importance of various factors affecting the extent of
company ownership, we are able to control for potentially unobserved firm-specific effects that
may cause patterns to differ cross-sectionally as opposed to over time within firms.
We use non-parametric estimation techniques and spline functions to model the evolution
of company ownership over time within franchised firms. We find that a significant decrease in
company ownership occurs during the first few years after a firm becomes involved in
franchising. This is consistent with empirical results currently in the literature (e.g. Gallini and
Lutz (1992), Lafontaine (1992), Scott (1995), and Thompson (1994)). However, we show that
the reason for this decline is quite simple: company ownership must fall in the first few years
because it begins, for all firms that are just getting involved in franchising, at 100%. We proceed
to demonstrate that after some initial years of decline, the percent of total outlets that are
company-owned is remarkably constant within firms over time. We conclude from this that dual
distribution is an equilibrium strategy for these firms, one that they adopt as they balance
incentive and control issues, or, put differently, to balance the benefits and costs of company
ownership and franchising. 2 Finally, we show that the stable level of company ownership varies
significantly across sectors, but also across firms within sectors. We find a strong positive
relationship between the “targeted” proportion of company ownership chosen by firms and
2 See e.g. Bradach and Eccles (1987) and Bradach (1997) for synergy arguments suggesting that dual distribution, orcontract mixing, is optimal in franchising.
4
measures of the value of their brand. We argue that firms with strong, high-quality brandname
strategies adopt higher targets of company ownership because they need to exert a greater degree
of direct managerial control in their chain.
The paper proceeds as follows. In the next section, we very briefly highlight some of the
theoretical arguments leading to predictions about how the proportion of company units should
evolve within firms over time. Following that, we describe our data and introduce non-
parametric and regression results that describe the within-firm evolution of the company
ownership of outlets. Given our results showing the stability of company ownership
percentages, we examine in Section 3 how firms successfully maintain these constant
proportions. We analyze how and why the extent of company ownership varies across firms in
Section 4. We draw implications from our results and conclude in Section 5.
2. COMPANY OWNERSHIP WITHIN FIRMS OVER TIME
2.1 A BRIEF OVERVIEW OF THE LITERATURE
Several existing theoretical explanations for franchising imply that the extent of company
ownership in these chains should change systematically as they become better established. Some
models, for example, assume that franchising is more profitable than company ownership, but
that firms must operate some units directly either to signal the high quality of their business
format and distinguish themselves from the low-quality firms (Gallini and Lutz, 1992) or to
credibly commit to upholding the value of the brandname (e.g. Scott, 1995). In these models,
firms want to reduce their extent of company ownership over time as their quality level is
revealed, or as they “develop a reputation for maintaining a certain level of quality” (Scott, 1995,
p. 76). Other models of franchising, by contrast, assume instead that company ownership is the
preferred mode of operation, but that franchising allows firms to obtain access to some scarce
5
resource, either capital (Caves and Murphy, 1976), managerial talent (Norton, 1988), or local
information (Minkler, 1990). As they become established and thus able to access these resources
directly, firms in these models want to reduce their reliance on franchising. In other words, these
models predict that the proportion of company ownership in a chain will grow over time.
Finally, most agency models of franchising have little to say about the extent of company
ownership in these chains as they focus instead on the optimal sharing rules and imply that the
terms of the franchise contracts should vary across units. Assuming, however, that franchisors
use only two types of contracts, a franchising and a managerial employment contract, then these
models imply that those units where high-powered incentives for the franchisees are very
valuable will be franchised, while others will be operated under a managerial employment
contract. The extent of company ownership in a chain then is a passive outcome arising from
outlet and franchisee heterogeneity in these models, and there is no expectation that the
proportion of company ownership should do anything systematic as firms become established.
(See Lafontaine and Slade, 2001, for more on this.)
A number of authors, including Gallini and Lutz (1992), Lafontaine (1992), Martin
(1988), Minkler and Park (1994), Scott (1995), and Thompson (1994), have examined the
relationship between the proportion of company outlets and franchisor age. Using cross-
sectional data, most have found evidence of a systematic decrease in the proportion of company
ownership as firms become established. Gallini and Lutz (1992) and Scott (1995) interpret these
results as evidence supporting the signaling or reputation effects described above.
2.2 THE DATA
Our extensive longitudinal data set allows us to revisit this issue empirically in a much
more precise way. We have information from 1980 through 1997 for more than 1000 franchisors
6
per year, resulting in a total sample size of 19162 observations. The data contain information on
1) the number of company-owned and franchised outlets, 2) the years of business and franchising
experience, 3) the royalty rates, advertising fees and franchise fees, and 4) a set of variables
describing some features of the franchisor, such as the amount of capital required to open an
outlet, and the type of business it is involved in. Our data set is a very unbalanced panel.
Though the number of franchisors included each year is rather stable, those firms that appear are
not the same from one year to the next. We have a total of 4842 different franchisors in our data
set, and so only about 4 observations per firm on average.3
Table 1 gives descriptive statistics for the main variables of interest for our overall
sample of 19162 observations.
TABLE 1: DESCRIPTIVE STATISTICS
Variable Name Mean StandardDeviation
Minimum Maximum
Number of Franchised Units 172.64 638.92 0 15394
Number of Company Units 35.05 234.05 0 4982
Total Outlets 207.69 788.15 1 19599
Percent Company Owned .222 .287 0 1
Years of Franchising Exp. 10.56 9.84 1 77
Years in Business 17.33 15.21 1 175
2.3 CHANGES IN THE PERCENT OF COMPANY OWNERSHIP AS FIRMS GAIN
EXPERIENCE: CROSS-SECTIONAL RESULTS
As a first step towards examining the evolution of vertical integration in franchised
chains, we show in Figure 1 the mean value of the percent company owned as a function of years
of franchising experience. We do this in Figure 1a for the overall sample, and then separately for
3 See the Data Appendix for more on the sources of the data and a detailed breakdown of the number of firms weobserve over different time frames.
7
chains with more than and less than the average number of outlets. In Figure 1b, we divide the
sample based on whether the firm is about to exit franchising or not, whereas in Figure 1c and
1d, we focus on different growth rates and different sectors respectively. Each time, we use all
the data and treat each observation as a separate piece of information. In other words, we do not
distinguish between cross-sectional or longitudinal patterns, and so the figure is dominated by
the cross-sectional effects.
FIGURE 1A: THE PROPORTION OF COMPANY UNITS AS A
FUNCTION OF FIRM EXPERIENCE AND CHAIN SIZE
2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 & up
0
10
20
30
40
50
60
Ave
rage
Per
cent
Com
pany
Ow
ned
Years of Franchising Experience
Full Sample
above mean (207.7)
below mean (207.7)
FIGURE 1B: THE PROPORTION OF COMPANY UNITS AS A
FUNCTION OF FIRM EXPERIENCE AND EXIT
2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 & up
0
5
10
15
20
25
30
35
40
45
50
Ave
rage
Per
cent
Com
pany
Ow
ned
Years of Franchising Experience
Full Sample
farFromExit
nearExit
8
FIGURE 1C: THE PROPORTION OF COMPANY UNITS AS A
FUNCTION OF FIRM EXPERIENCE AND GROWTH
2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 & up0
10
20
30
40
50
60
70
Ave
rage
Per
cent
Com
pany
Ow
ned
Years of Franchising Experience
Full Sample
posGrowth
negGrowth
zeroGrowth
FIGURE 1D: THE PROPORTION OF COMPANY UNITS
AS A FUNCTION OF FIRM EXPERIENCE AND BUSINESS ACTIVITY
2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 & up0
10
20
30
40
50
60
Ave
rage
Per
cent
Com
pany
Ow
ned
Years of Franchising Experience
automotive
business services
restaurants
construction and maintenance
retail
There are three striking features in Figure 1. First, no matter what sample one focuses
on, the percent of company ownership falls precipitously during the first eight or so years in
franchising. This is not surprising: Firms almost always begin franchising after having opened a
few company-owned outlets, and so they are almost always 100% company owned initially.
Thus, this proportion must fall in the early years in franchising, and we see that indeed it does
9
fall sharply. Note that this pattern is the source of the negative relationship between franchising
experience and company ownership that has been picked up to date in cross-sectional analyses.4
The second striking feature in Figure 1 is that after this initial decline, the percent company
owned stabilizes in our overall sample and most sub-samples at a surprisingly constant rate of
15% on average. Thus, we see signs that firms are holding their proportion of company
ownership quite fixed once they have reached a “target” level – which, on average, is clearly
neither 0 nor 100%. Finally, we see that the “stable” level of company ownership is the same for
firms of differing success levels as captured by the size of the chains, their “near exit” status or
not, and their growth rates. The one dimension along which we find different “targets” of
company ownership is the firm’s sectors of operation. In fact, we picked the sectors depicted in
Figure 1d specifically to highlight the differences in the “targets.” The figure clearly shows that
restaurant chains stabilize at a relatively high proportion of company units, while the
construction and maintenance sectors show a very low “stable” reliance on company ownership.
2.4 CHANGES IN THE PERCENT OF COMPANY OWNERSHIP AS FIRMS GAIN
EXPERIENCE: L ONGITUDINAL RESULTS
Figure 1 displayed cross-sectional patterns in the extent of company ownership, and
showed, for example, that all the firms that have 10 years of franchising experience have an
average percentage of company-owned units of 15 % and that firms that have 11 years of
experience also have 15 percent company units. But these mean values could hide important
within-firm changes over time. For example, the constancy of the mean value of the percent
4 Consistent with these cross-sectional analyses, if we fit a regression to our overall sample with a quadratic, cubeand quartic term, and even higher order terms, we always find that the first-order term in the regression is large andnegative. Furthermore, we find all higher-order terms to be significant, despite the fact that the raw data reveal aclear lack of systematic pattern after the initial period of decline. In other words, those various terms are picking upsmall variation in the data and given our very large sample, each of these is statistically significant.
10
company owned could mask a decline for some firms that is offset by an increase for others. In
this section, we model the within-firm evolution of ownership, namely
∆Cit = f(franchising experience)it + eit (1)
where ∆Cit is the within-firm change in the percent company owned for firm i at time t, which
varies as a function, f, of franchising experience. The residuals eit represent unexplained
variation in ownership patterns. In what follows, the function ‘f’ is estimated first using non-
parametric, and then with parametric methods.
2.4.1 Non-parametric Results
Using Kernel regression, we show in Figure 2 how the change in company ownership
within firms relates to franchising experience. The graph shows results obtained with different
bandwidths, from two to five years of experience, and how the results compare to a straight line
at 0. Though this is not easy to perceive from the graph, as one would expect, the results become
smoother as the bandwidth widens.5
What is most remarkable about Figure 2 is that the change in the percentage of company-
owned units flattens out at zero at about 8 years of franchising experience. In other words, from
that point on, the average firm makes no more changes in its proportion of company units. This
remains true when we look at sub-samples of the data as we did before, namely for firms with
positive, negative, and zero rates of growth, or for firms with above and below average numbers
of outlets, or for firms within 3 years of exiting or not, or for firms in the five sectors defined
earlier. In all cases, when looking at changes in the percentage of company ownership as firms
gain experience, the change settles at zero after about 8 years of franchising experience.
5 See Salgado-Ugarte, Shimizu and Taniuchi (1993) for more on bandwidth or “smoothing parameter” choices.
11
FIGURE 2: KERNEL RESULTS FOR THE CHANGE IN PERCENT COMPANY OWNED
FOR DIFFERENT BANDWIDTHS (2, 3, 4, 5)
Years of Franchising Experience0 5 10 15 20 25 30 35 40
-12
-10
-8
-6
-4
-2
0
4.2.2 Parametric Regression Results
We use the patterns evident in the graphs above to construct reasonable parametric spline
functions that model the change in percent company-owned units as firms gain experience. The
chief advantage of the parametric approach is that we can easily devise statistical tests to answer
two questions: 1) do the percentages of company ownership really stabilize over time, as they
seem to? And 2) are there significant differences across sectors or other sub-samples of firms in
the extent to which they stabilize?
The regression model that we estimate, and whose results are shown in Table 2,
introduces a simple linear coefficient on experience in the regression of the change in percent
company-owned of (1), but this linear coefficient is allowed to vary by experience level. We
permit three intervals for franchising experience: 1-3, 4-7, and 8 to 45 (the sample is truncated at
46 years of experience due to insufficient data beyond that).
12
TABLE 2: SPLINE REGRESSION RESULTS FOR THE CHANGE IN THE
PROPORTION COMPANY OWNED
Sample Number ofObs.
Years of Franchising Experience
1-3 4-7 8-45
FULL SAMPLE 13022 6.42 (24.81)
.431(4.92)
-.005(-0.33)
GROWTH RATES:
Positive 9959 6.69(24.22)
.588(6.00)
.013(0.64)
Negative 3063 1.24(1.67)
-.505(-2.59)
-.045(-1.62)
Zero 2102 1.68(3.46)
.403(2.19)
.001(0.02)
NEAR EXIT STATUS:
Yes 4609 6.26(13.46)
.371(2.26)
.006(0.18)
No 8413 6.58(20.93)
.465(4.58)
-.012(-0.63)
NUMBER OF UNITS:
Above Average 2579 1.62(3.12)
.435(4.19)
-.002(-0.21)
Below Average 10443 6.51(22.38)
.439(4.19)
-.007(-0.26)
SECTOR:
Automotive 1072 5.65(6.29)
.614(2.10)
-.011(-0.20)
Business Services 1592 3.09(4.28)
.468(1.89)
-.004(-0.09)
Restaurants 3466 4.29(8.03)
.734(4.32)
-.004(-0.16)
Construction & Maintenance 1199 8.42(9.85)
.247(0.86)
.005(0.08)
Retail 2599 9.46(16.61)
-.082(-0.41)
.023(0.50)
Note: t-statistics in parentheses.
The spline regression results in Table 2 reveal that after the first seven years of
franchising experience, there is no significant variation in the percent of company ownership
over time within firms. Furthermore, F-tests show that after the first seven years of experience,
there are no significant differences in data patterns across sectors or for different sub-samples.
13
The stabilization of the percentage of company-owned units occurs among all these sets of firms.
Finally, we tested a large number of alternative spline functions, all of which were rejected by
the data. For example, if we permit the linear relationship to vary within the experience levels 8
through 45, we find no significant effects. Thus, the data imply that there is no need for more
cut-offs within this period.
3. HOW DO FIRMS M AINTAIN THIS CONSTANT PROPORTION OFCOMPANY OWNERSHIP?
The results above show that the percentage of company ownership remains fixed after
the early years of franchising. In this section, we examine how this is achieved. In other words,
we ask whether we see this constancy because many firms are simply not expanding after their
first few years in franchising, or whether this constancy occurs because firms make important
changes in their number of company owned units as they grow or shrink the franchise side of
their business.
In Figure 3a, we show the mean changes in the number of company and of franchised
outlets respectively as firms gain experience. Concentrating on firms that are growing with age,
we see that the change in the number of company-owned outlets rises over time. On average,
during experience years 1 through 8, one or two company-owned outlets are added per year;
during experience years 9 through 18, three to five company-owned outlets are added on average
per year; and then for experience levels above that, ten to fifteen company outlets are added per
year. Figure 3b shows the accompanying change in the numbers of franchised outlets. Mostly,
the combination of the two figures illustrates how firms, when growing or shrinking, change both
their number of company owned and of franchised units significantly in the process of
maintaining their constant overall proportion of company units.
14
FIGURE 3A: THE CHANGE IN THE NUMBER OF COMPANY UNITS BY EXPERIENCE LEVEL
2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 & up-15
-10
-5
0
5
10
15
20
25
Ave
rage
Cha
nge
in C
ompa
ny U
nits
Years of Franchising Experience
Full Sample
Positive Growth Subsample
Negative Growth Subsample
Zero Growth Subsample
FIGURE 3B: THE CHANGE IN THE NUMBER OF FRANCHISED UNITS BY EXPERIENCE LEVEL
2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 & up-60
-40
-20
0
20
40
60
80
100
Cha
nge
in F
ranc
hise
d U
nits
Years of Franchising Experience
Full Sample
Positive Growth Subsample
Negative Growth Subsample
Zero Growth Subsample
We conclude that the “stable” percentage of company units found above is not an artifact
of lack of growth, but rather is achieved by firms via active management. In other words, when
they add new units, they purposely add both company units as well as franchised ones to
maintain their overall target of company-ownership. Thus, the proportion of company-owned
15
units in these chains is not simply a passive outcome, which would arise simply from the
combination of unit-level market characteristics and the need to choose the right contract for
each unit, as agency theories would suggest. Rather, it is a decision variable that firms actively
work toward. (See also Scott, 1995 on this.) On the other hand, these firms work towards a
stable proportion that is neither 0 nor 100%, thereby contradicting other theories that suggest
they should be aiming for fully franchised or fully company-owned structures.
It is important to note that this conclusion, that firms target certain levels of company
ownership, is consistent with companies’ stated strategy, as reported in various sources. For
example, Howard (1994), in his article on Burger King, refers to “the 7432-unit chain’s goal to
eventually own and operate just 10 percent of the system.” Similarly, Sasser and Pettway (1978)
point out that “Wendy’s expansion strategy was based on a “balanced” development of
company-owned and franchised stores” (p. 104) and “Waffle House was expanding on two major
fronts: company-operated units, and franchised units. (…) It was expected that an eventual
balance of the system would be reached at 30 percent company stores and 70 percent franchise
stores.” (p. 139). Furthermore, the notion that firms set a goal for their proportion of company
units is supported by the fact that franchisors, when asked, will answer questions concerning the
“optimal” proportion of company units in a chain. Lafontaine and Kaufmann (1994) analyze
survey data on this, and find that in most franchising firms, respondents believe that they operate
what they perceive to be the optimal proportion of company units. The authors also note that in
the vast majority of cases, this proportion is not 0 nor 100%.
In its 1995 annual report, McDonald’s gives reasons why it wants to operate units
corporately. It states that “About 21% of McDonald’s restaurants are operated by the Company.
We maintain a global base of Company-operated restaurants to generate profits, link our interests
with franchisees, develop management talent, gather research, and test ideas for better restaurant
16
execution” (p. 11). We explore some of these arguments in our empirical analyses of the target
level of company ownership in the next section.
Before turning to such analyses, we want to point out that our conclusion that franchisors
choose a particular level of company ownership does not imply that they cannot also choose
which units should be franchised and which should be company operated. Moreover, they can
very well base this particular decision on local market and unit characteristics. Existing evidence
strongly suggests that smaller, rural units that are farther away from headquarters are more likely
than others to be franchised (see Lafontaine and Slade (1997, 2001) and the references therein).
Thus local market and unit characteristics clearly affect the ownership decision. However, our
results imply that a conscious decision to operate a certain proportion of units does inform or
constrain the unit-level decisions on ownership.
4. EXPLAINING THE DIFFERENCES IN TARGET LEVELS OF COMPANYOWNERSHIP
Thus far, we have shown that once franchisors become established, they target a constant
level of company ownership—that is, the percent of outlets that are company owned is
remarkably stable within firms over time. However, the target levels vary significantly across
firms. The firm-specific nature of the company ownership targets can best be shown empirically
as follows: if we restrict our sample to established franchisors (those with more than 7 years of
franchising experience and at least 15 units) and run a regression of the percent of company
ownership as a function of a set of explanatory variables and firm-specific fixed effects, the
introduction of the firm-specific fixed effects raises the R-squared from 18.4% to 90.2%.6 Thus,
6 The explanatory variables in the regression are those introduced in column 7 of Table 5. They include 22 sectordummies that contribute 7.5 percentage points to the R-squared.
17
it is clear from the data that different franchisors adopt strategies involving different targets. In
this section, we address the question: What determines the target levels of company ownership
that companies choose?
4.1 THEORETICAL ARGUMENTS
As noted earlier, basic incentive-cum-insurance models of franchising imply that the
proportion of company units is a passive outcome that depends only on the characteristics of the
set of outlets in the chain. Other models, such as signaling and reputation-based models imply
systematic patterns up or down as firms become established, not the kind of stable “target” levels
of company ownership that we observe. Thus, we must turn elsewhere for explanations of stable
“target” levels of company ownership.
There are two main types of arguments in the literature that can explain differences in the
extent of company ownership across firms and suggest factors that can influence this choice.
The first relates to franchisee free-riding and the resulting need for the franchisor to control
downstream operations. The second relates to franchisees’ need to insure that their franchisors
will not behave opportunistically once the contract is in place, namely that the franchisor will not
under-invest in the brand. We briefly detail each of these and their empirical implications in the
following section. Before turning to these, however, we note that several other models have
been proposed to explain contract mixing. For example, Bai and Tao (2000) use a multi-tasking
approach to argue that franchisors offer low-powered incentives (a company manager contract)
to agents who are called upon to put effort in the generation of goodwill, and high-powered
incentives (a franchise contract) to those who will put effort into sales generation. One
implication of their model is that franchisors that require more goodwill effort will have to
operate more company units. That is, unless the franchisor itself provides the goodwill effort, as
18
is argued elsewhere in the literature. Other work by Bradach (1997) and Lewin (1998)
emphasizes the synergies between company units and franchised units. For example,
McDonald’s affirms that it operates units directly in part to “gather research, and test ideas for
better restaurant execution.” In fact, franchisors not only can test ideas in company units, they
can also use results from these units to demonstrate the profitability or feasibility of various ideas
and innovations. In that sense, company units serve the roles of both information generation and
diffusion within the franchised system. As suggested by the “window on the industry argument,”
company units also provide franchisors with information about individual markets, information
that they may have more difficulty obtaining from other franchisees and that they can use to
assess the behavior of individual franchisees. While all these models are of interest, they provide
little guidance as to what the optimal mix of units is, and what it should relate to empirically. For
that reason, we focus on the models below.
4.1.1 Franchisor’s Need for Control
When firms choose to franchise, they give residual claimancy rights (after payment of
royalties on sales) to their franchisees. This gives franchisees good incentives to keep costs
down and sales up. Unfortunately, as pointed out e.g. by Brickley and Dark (1987) and Blair and
Kaserman (1994), the residual claimancy status of the franchisees also gives them incentives to
maximize their profits sometimes at the expense of the chain. For example, franchisees can
choose to use lower quality inputs or offer less services to reduce their costs – they obtain the full
increase in profits that results from these actions, but bear only part of the consequences of lower
brandname value. Furthermore, franchisees may adapt their product offerings to better suit their
local markets in ways that affect the value of the brand. In either case, franchisors will want to
exert control and prevent these behaviors.
19
Following Blair and Kaserman (1994), we illustrate the free-riding problem formally as
follows. Define p={p1, p2, …pn} and s={s1, s2, … sn} to be the price and service quality levels
chosen by the n franchisees in the chain. Assume that every franchisee’s profit level, πi, is a
function of these two vectors, with ∂ πi/∂pj < 0 and ∂ πi/∂sj > 0.7 For simplicity, assume that the
franchisor’s profits, Π, is derived as a simple share of each franchisee’s profits, namely
Π ==∑δ π( ( , ))ii
n
p s1
.8
If δ=1, Π is total chain profits. To maximize the franchisor’s (or chain’s) profits, the service
level chosen by franchisee j would have to satisfy
∂∂
= ∂∂
==∑Π
s sj
i
ji
n
δ π
1
0
which, given the constancy of δ, can be rewritten as
∂∂
= − ∂∂=
≠
∑π πj
j
i
jii j
n
s s1
.
This level of franchisee service maximizes the franchisor’s (or the chain’s) profits, that is
it takes into account the demand externalities and acts upon them. However, the individual
franchisee, as a residual claimant, will want to maximize its unit’s profits. This requires setting
∂ πj/∂sj = 0, which, of course, implies a lower service level than the one that maximizes the chain
7 Note that intra-chain competition could imply ∂ πi/∂pj > 0 and ∂ πi/∂sj < 0. While units may compete with a few
other units in their chain, for the majority of cases, the interdependence due to the shared brand dominates.8 We make the assumption that the franchisor’s profits relates to the franchisee’s in a simple way for simplicity. Ifwe assumed instead that the franchisee pays a portion of its sales to the franchisor, the qualitative results would beunchanged. Similarly, an assumption that different units contribute differently to the franchisor’s profits would notchange the qualitative result of under-investment in activities where there are externalities.
20
profits. Hence, from the overall chain’s perspective, the franchisee under-invests in service
provision activities. Similarly, the franchisee will under-invest when it comes to its pricing
decisions, namely he will choose a price that is too high from the chain’s perspective.
The issue, as just shown, arises from the fact that each franchisee’s profits or demand
level (and hence total demand for the chain) depends on decisions made by others in the chain.
The stronger this dependence is (i.e. the larger is ∂ πi/∂sj), the more franchisee choices will
diverge from those that would maximize the chain’s profits. Yet franchising, by its nature,
implies interdependence. In fact, this interdependence, which occurs through the brand, is a
large part of what franchisees buy when they purchase a franchise. The more valuable the brand
is, the more franchisees profit from it, and the more their revenues and profits are interrelated.
Put differently, the more valuable the brand, the higher ∂ πi/∂sj and ∂ πi/∂pj will be. This, in turn,
will lead to greater divergence between the service and price levels that franchisees choose
compared to the levels that the franchisor wants them to use.
Note that if franchisees were not residual claimants, but were paid a fixed salary as
corporate managers are, they would have no incentive to choose a different price and service
level than the one requested by the franchisor.9 Thus, the divergence in prices and service levels
amounts to a cost of using high-powered incentives for franchisees. This divergence, and hence
the cost of using high-powered incentives, will increase with the extent of interdependence in the
chain, or with the value of the brand.
9 This assumes that they do not bear the cost of providing the requested service level. If they do, they will haveincentives to shirk on service quality. This is the incentive problem that the franchisee’s residual claimancy status ismeant to address in the first place. We assume that franchisors can use other mechanisms to prevent shirking incompany stores at least to some degree.
21
The simple model above emphasizes differences in quality and price levels chosen by
franchisees compared to what is optimal for the chain. This model thereby illustrates the
standard “free-riding” argument against franchising. But as mentioned above, profit-maximizing
franchisees may also rationally oppose the implementation of new production processes and new
product offerings that they expect will not do well in their particular market - though the new
method may be right for most other stores in the chain. Franchisees might also introduce
products that are suited to their market, even though they are not suited to the chain as a whole.
Such behaviors also hurt the value of the brand as branding is fundamentally linked to
uniformity, in product offerings, service levels, prices, etc. These behaviors also are caused by
the franchisee’s residual claimant status, and as such represent a cost of franchising, a cost that
grows with the value of the brand and the resulting need for consistency and uniformity.
Franchisors can impose various contractual restraints to curb the behaviors just described,
such as requiring some minimum advertising and service levels, or imposing maximum prices, or
by developing strict product mix guidelines.10 But such restraints are costly to enforce, and more
to the point, may not prevent all the types of behaviors franchisors care about. To protect their
brand further franchisors may need to exercise more direct managerial control over downstream
operations. Vertically integrating with some of their units will offer franchisors the opportunity
to control the behaviors just described directly in the integrated units, and indirectly in the
franchised units. The indirect control arises from competition among corporate and franchised
units, from the capacity to demonstrate the feasibility and profitability of new policies and new
10 A 1997 court decision, State Oil v. Khan, has moved maximum resale prices to the rule of reason approach,making them for all intents and purposes legal. However, until this time, the imposition of maximum prices byfranchisors was per se illegal, as were all forms of resale price maintenance.
22
products, and, finally, from the market intelligence garnered from operating units directly.11
We conclude that whether it be in the form of the usual free-riding issue, or in the form of
resisting change or adapting too much to local markets, the cost of franchisees maximizing their
own unit’s profits at the expense of those of the chain increases with the value of the brand.
Ultimately, when this cost becomes large enough, firms will have a greater tendency to exert
direct managerial control over downstream operations. Thus, we expect that the more valuable
the brand, the higher the extent of company ownership should be.
4.1.2 Aligning Franchisor Incentives
A number of authors have emphasized the role of the franchisor in the ongoing value of
the franchised business: the franchisor puts forth effort in advertising and promoting the brand,
in monitoring franchisees, and in screening new franchisees, all of which are necessary for the
long-term profitability of the chain. Models of franchising that have explicitly incorporated the
franchisor’s role (and the difficulty inherent in motivating his behavior) have shown that output
sharing will be necessary to motivate the franchisor – or, as stated by McDonald’s, to “link the
franchisor’s interests with those of its franchisees.” This can be achieved, in a single franchisor-
franchisee pair, with an appropriate choice of royalty rate (see e.g. Rubin, 1978, Lal, 1990, and
Bhattacharyya and Lafontaine, 1995 for formal models). In that case, the more important is the
effort of the franchisor in the production process, and the more difficult his role is to monitor, the
more the franchisor should be given incentives, so the higher the royalty rate should be.
As noted in Lafontaine (1992, 1993) and Scott (1995), an alternative way to give the
franchisor a stake in downstream operations, and thus incentives to maintain the value of the
brand, is to have him operate units directly. If this is the mechanism by which the franchisor is
11 This last aspect amounts to what is known as the “window on the industry” argument.
23
given incentives, then the proportion of company units should be higher the more important the
effort of the franchisor and the more difficult it is to monitor. Under the assumption that the role
of the franchisor is to develop and uphold the value of the brand, the proportion of company
units should increase with the importance of the brand in the success of the business.12
Finally, while agency arguments emphasize the incentive effects of the franchisor’s stake
or residual claims in the business, everything else constant, a larger proportion of franchisor-
owned units in the chain also gives franchisors incentives to perform if it means that they have
more specific assets tied into the system. The need to protect these assets will motivate the
franchisor to maintain the value of the brand and of the franchised system as a whole.
4.2 THE MODEL AND DATA FOR TARGETING ANALYSES
The implication from both sets of arguments above is that in equilibrium, that is once
they are established, franchisors will choose to operate more units directly the more valuable
their brand is. Our primary hypothesis then is that:
Hypothesis: Franchising firms with higher tradename value will target a higher level of
company ownership.
We test this hypothesis by estimating the following regression:
Pit = aXit + bZit + eit (2)
where Pit is the percentage of company-owned outlets by firm i at time t, Xit is a vector of
variables representing possible tradename value proxies, and Zit is a set of controls, such as
sector dummies. The specification for the residual eit is described below.
12 Franchisees indeed worry about large changes in this proportion (Bond, 1997). If the franchisor reduces itsinvolvement downstream importantly, franchisees worry that he no longer will invest. Increases in this involvementcan be interpreted as an increased probability that the franchisor is trying to take over downstream operations in anopportunistic way. Either way, change is worrisome.
24
We explore the relationship between the target level of company ownership and the value
of the tradename by restricting our sample to those firms and observations with 8 or more years
of franchising, and 15 or more units. We do this because we wish to examine only those years
when companies have grown enough to have reached their target levels—in the early one to
seven years of franchising, the development of the first few franchised outlets must produce a
drop in the percent company owned, as shown in Section 2 above. Table 3 shows descriptive
statistics for this new “stable” sample.
We use three primary measures to assess the value of the brand: 1) the advertising fee,
which is the % of their outlet sales that franchisees are required to pay to their franchisor for
advertising purposes (Advertising), 2) the number of years the franchisor spent in business
developing the system before beginning to franchise (Years Before Franchising), and 3) the
amount of advertising dollars spent across major media for the brand (Media Exp.). The first
two measures were collected from the same sources as all our other franchising data. However,
they are not ideal proxies for tradename value, in that the advertising fee data is somewhat
incomplete 13, and Years Before Franchising could raise company ownership simply because a
large volume of outlets are initially company owned. Thus, we sought a more accurate measure
of tradename value, and collected the yearly media advertising expenditures reported in $Ad
Summary for each brand in our “stable” sample from 1980 to 1997. (Other variables in Table 3
are described in Section 4.3).
13 Our measure of advertising fees is the extra fee that some franchisees must pay to franchisors for advertisingexpenses. Some franchising firms (many in the early years of our data) absorb this fee in the required royalty ratesthat they charge franchisees and do not charge a separate advertising fee. Thus, our measure of advertising feeunderestimates investments in advertising for an unknown number of firms.
25
TABLE 3: DESCRIPTIVE STATISTICS FOR THE “STABLE” SAMPLE
Variable N Mean Std. Dev. Min Max
% company owned 7906 .134 .200 0 1
Years Before(experiencebefore franchising)
7906 7.86 10.93 1 128
Number of Outlets 7906 396.49 1093.89 15 19599
Advertising (% of sales) 7906 1.57 1.87 0 13
Advertising /Outlet 7906 .021 .039 0 .469
Media Exp. (in $1000) 7906 2780.6 21445.2 0 598752.5
Media Exp./Outlet 7906 2.28 7.46 0 259.7
Products Only 7906 .516 .500 0 1
Business Only 7906 .091 .288 0 1
Capital Required ($) 7670 231.89 1472.23 0 82500
# of States franchising in 5950 19.73 16.22 0 50
Financing Available(from franchisor)
7779 .688 1.054 0 4
Training (in days) 5693 18.80 17.18 0 290
Experience Required (offranchisees)
4073 .271 .444 0 1
Contract Length (years) 3690 11.94 6.162 1 50
# of Employees (peroutlet)
3547 9.21 13.48 0 150
4.3 SEMI -PARAMETRIC REGRESSION RESULTS
We test our primary Hypothesis—that high tradename value increases company
ownership targets—by estimating regression (2) using several alternative functional forms.
Results are shown in Table 4, using three different estimation techniques—OLS, Tobit, and a
Tobit between estimator. As a point of comparison, we first present OLS results obtained using
the whole “stable” sample. Then, in recognition of the fact that we have 26.2% of our PCENTC
forming a mass point at zero in the “stable” sample, we present results obtained with a Tobit
estimator. Finally, for the remainder of the regression results, we move to a between Tobit
estimator, where we use firm-specific averages as data points and weigh each observation
according to the number of observations per firm that is used to calculate the firm-specific
26
averages.14 We switch to the between estimator for one important reason—we have shown
above that firms change their targeted level of company ownership very little over time. Thus
using the full stacked data set can amount to the redundant use of multiple observations per
company. Instead, by using the between estimator, we are emphasizing how tradename value
matters as an explanation of the variation across firms in the data. This has the added advantage
of “averaging” all our explanatory variables over time as well, making our media expenditure
variable a less volatile and more representative measure of long-term investment in brand value.
The results in Table 4 show that tradename value, per all three of our measures, has a
very significant positive effect on the extent of company ownership. Our three measures
combined explain 13 percent of the variance of company ownership across firms, and the
Media/outlet and Advertising/outlet explain 11 percent of this variance.
The most important effect arises from the value of Media Expenditures per outlet, where
we have divided Media Expenditures by the number of outlets because large firms are likely to
spend more on advertising than are small firms.15 The Media/outlet effect is estimated in a semi-
parametric quartic specification, given its theoretical importance and the statistical significance
of all four non-linear terms—thus, the effect of Media/outlet rises, then plateaus, then rises again
at substantially higher levels.16 The Media/outlet coefficients in the between-Tobit regressions
imply a sizable impact: at the mean value of Media/outlet, a 10 percent increase in expenditures
14 The typical between estimator does not use firm-specific weights reflecting the number of observations per firm,but between estimators are normally used for panel data that are balanced, thus this issue would not arise. For ourunbalanced panel data set, measurement error is lower for these firms with more time-series data, and thus theseshould be more heavily weighted.15 If Media is included instead of Media/outlet, the significance of Media is essentially the same as Media/outlet.16 The effects of Media/outlet was also estimated in numerous other functional forms and the quartic was found to besuperior. For example, when estimated in a spline function with seven break points in Media/outlet, the pattern inthe coefficients is still quartic and there is no significant gain from the less constrained spline function relative to thequartic. We also plotted the predicted value of the effect of Media/outlet on PCENTC and find that the combined
27
will create a 9.6 percent increase in company ownership, or increase the company ownership rate
by .7 percentage point.
While Media/outlet has the predominant tradename value effect in our regressions, two
other variables measuring tradename value are also highly significant. The variable
Advertising/outlet, has a strong positive linear effect (squared terms are insignificant) across all
estimated functional forms.17 Similarly, firms that developed more experience and thus higher
tradename value prior to franchising (as measured by number of years in business before
franchising) are more likely to have higher levels of company ownership. Of course, it is also
the case for these firms that they were more likely to have begun franchising with a more sizable
number of company-owned outlets.
In these regressions, we introduce several sets of controls, for survey year (dummies for
years 1980-1997), vintage (dummies for year in which the firm began franchising), and company
size (total number of outlets). The vintage dummies show a significant pattern of rising
company ownership rates.18 This result implies that more recent entrants into franchising have
been targeting higher proportions of company ownership on average. This effect possibly
reflects the increased maturity of industries in which franchisors operate, and the resulting
increase in the importance of brandname value for new entrants in these industries. Turning to
the company size variable, the number of outlets has a negative effect on company ownership in
the basic Tobit model, but this effect disappears in models that control for sectors as well as in
linear-squared-cubic-quartic effects have sensible turning points, but the levels overshoot somewhat.17 The advertising variable is the advertising fee paid as a percentage of outlet sales, and it performs best in ourregressions when it is divided by the number of outlets, suggesting that the advertising fee has a smaller impact oncompany ownership rates for large franchisors.18 The coefficient (t-statistic) on the linear year-started-franchising is .0014 (4.60) if this variable replacesfranchising vintage dummies in column 2 of Table 4. The franchising entry-year dummies are segmented into eightdummies as follows: 1927-59, 1960-64, 1965-69, 1970-74, 1975-79, 1980-84, 1985-89, 1990-97.
28
the between models, suggesting again that there is no change in company ownership rates as
firms age and grow.
TABLE 4: PERCENT COMPANY OWNED REGRESSIONS
OLS Tobit Tobit TobitBetween
TobitBetween
MEDIA/OUTLET 2.30(21.90)
2.91 (21.79)
2.36 (17.77)
3.41 (9.44)
2.72 (8.11)
(MEDIA/OUTLET)2 -.066 (14.06)
-.085 (-14.32)
-.069 (-11.80)
-.102 (-5.87)
-.086 (-5.07)
(MEDIA/OUTLET)3 .0006 (11.04)
.0007 (11.42)
.0006 (9.48)
.0009 (4.21)
.0008 (3.95)
(MEDIA/OUTLET)4 -.000001(-9.79)
-.000002 (-10.21)
-.000001 (-8.51)
-.000002 (-3.62)
-.000002 (-3.54)
ADVERTISING/ OUTLET
.364 (6.36)
.499 (6.85)
.325 (4.40)
.788 (5.82)
.351 (2.20)
YEARS BEFORE .393 (10.11)
.524 (10.58)
.541 (11.29)
.639 (6.73)
.456 (4.98)
(YEARS BEFORE)2 -.0022 (-3.51)
-.0034 (-4.41)
-.0035 (-4.65)
-.0056 (-4.28)
-.0029 (-2.01)
OUTLETS -.102(-2.26)
-.121(-2.11)
.006(0.10)
-.268(-1.52)
-.075(-0.55)
(OUTLETS)2 .0010(2.19)
.0011(2.08)
.0002(0.45)
.0028(1.49)
.0011(0.74)
SECTOR DUMMIES No No Yes No Yes
χ2 χ2(34) =
1048.54
χ2(56) =
1714.96
χ2(34) =
266.77
χ2(56) =
383.31
R2 .123
N 7906 7906 7906 1563 1563
t-statistics are in parentheses. Also in regression: Canadian dummy, year dummies (for 1980-1997), franchisingentry-year dummies (see footnote 18). The full sample Tobits have 2070 observations censored at zero, and 5836uncensored.The between sample Tobits have 266 observations censored at zero, and 1297 uncensored.
When we add sector effects to the regression, they are very significant as a group,
suggesting that there are differences in the desirability of franchising across the different types of
business activities (compare columns 2 and 3, and then 4 and 5), but they do not alter our
conclusions on tradename value—tradename value increases company ownership within sectors
29
as well as across sectors.19 There are sector-specific differences in the rate of company
ownership, and these differences could be related to the different types of production functions
involved, which put more or less weight on franchisor and franchisee input, and imply different
levels of monitoring costs as well. We explore sectoral differences in more detail below.
In column 1 of Table 5, we replace the sector dummy variables with only two dummies,
one for firms producing products only (as opposed to products and services, or services only) and
one for those supplying only businesses (as opposed to individual consumers). Interestingly, we
find that franchisors that sell a product for resale tend to rely on company ownership more than
those that sell services. This is consistent with the theory that franchisors have more at stake and
want to exert more direct control when they sell their product through resellers under their brand.
Service franchised chains have considerably less company ownership on average—about 9%
less. Service firms, such as dry cleaners or tax preparation businesses, also depend on franchisee
behavior to maintain brandname value, but are on average likely to have less established
brandnames (our media variable has a smaller mean for services) and to face more repeat
business, which generates incentives at the local level for maintaining one’s own reputation.20
19 Specifically, the sector-specific dummy coefficients are as follows:Automotive products .082*** Real Estate -.096***Business Services .113*** Recreation .066Business Supplies -.058*** Rental .066Contractors -.020*** Repair .133Cosmetic Products & Services .119** Retail-Build Mat .091Eating places—Full Service .262*** Retail-Clothing .146***Eating places—Limited Service .158*** Retail-Food .164***Education -.012*** Retail-Furnishing .143***Health & Fitness .240*** Retail-Other .084**Hotels and motels .150*** Retail-Used .080Maintenance .015*** Travel .015***Personal Services .120***These represent partial effects—holding constant all other variables in column 2 of Table 4. *** is significant at the1% level, and ** at the 5% level.20 While these results are consistent with maintaining brandname value and minimizing externalities and free-ridingproblems, they contradict the idea that service production requires more effort incentives than product reselling.
30
TABLE 5: PERCENT COMPANY OWNED REGRESSIONS
1Between Tobit
2Quantile .33
3Quantile .5
4Quantile .67
MEDIA/OUTLET 3.14(9.50)
.933(17.96)
2.11(12.98)
4.22(10.25)
(MEDIA/OUTLET)2 -.100(-5.93)
-.0175(-7.22)
-.053(-6.75)
-.155(-7.99)
(MEDIA/OUTLET)3 .0009(4.66)
.00013(4.82)
.0004(4.43)
.0016(7.33)
(MEDIA/OUTLET)4 -.000002(-4.19)
-.0000003(-4.06)
-.0000009(-3.62)
-.0000043(-7.06)
ADVERTISING/OUTLET .493(3.21)
.324(14.60)
.635(9.58)
1.063(6.29)
YEARS BEFORE FRANCHISING .444(4.73)
.081(5.15)
.213(4.57)
.320(2.70)
(YEARS BEFORE FRANCHISING)2 -.003(-2.11)
-.0006(-2.92)
-.0021(-3.44)
.0009(0.53)
OUTLETS/1000 -.132(-0.97)
(OUTLETS)2 .0014(0.93)
PRODUCT SECTORS .077(7.03)
BUSINESS SECTORS .055(2.88)
SECTOR DUMMIES NO NO NO NO
χ2 χ2(36)=305.27
R2 .0528 .081 .120
t-statistics in parentheses. Regressions are comparable to those of column 4 of Table 4. Also, in the regressions are:Canadian dummy, year dummies (1980-1997), franchising entry-year dummies (see footnote 18). Sample size is1563 for these between regressions. The quartile regressions do not control for the masspoints at zero (they are notTobits).
In Table 5, we explore the sectoral differences in more detail by looking at quantile
regression analysis (columns 2-4). The firms in the upper two-thirds (or .67 quantile) of the
distribution of PCENTC clearly have bigger Media/outlet and Advertising/outlet effects than
those in the lower third quantile (columns 2 vs. 4).21 These differences across quantiles suggest
21 We also show the results for the median regression in column 3, which diminishes the effects of outliers on the
31
that firms with high tradename value, or Media expenditures, are targeting high company
ownership—yet within the sectors where we see high values of company ownership (as in the .67
quantile) there is still a big effect of Media on ownership.22
In Table 6, we add more explanatory variables to our Tobit regressions, and find that
these additional variables also strongly support the hypothesis of managerial targeting of
company ownership to control brandname value.23 To start with, in all our regressions, Training
Required has a very strong and significant positive effect on the proportion of company
units—every two days of additional training required by the franchisor raises the target
proportion of company units by almost 1%. Of course, firms that place a high value on their
brandname must require more training to maintain product quality in close line with company
standards. Furthermore, the provision of more training by a franchisor likely reflects a more
developed business format and thus stronger brand. In fact, training and media expenditures are
highly correlated in our data (a correlation of .52), though their effects are both positive and
jointly significant in the regressions.
Similarly, the result that franchisors with physically larger units (higher Capital Required
to purchase a franchise or higher Number of Employees per outlet) tend to own more units also
supports the tradename value argument: franchisors have more at stake with every unit in chains
where units are large, operations are more complex in such units, and they can generate higher
results relative to the standard Tobit regression, and we find no change in the coefficient estimates. Note also that ifwe drop McDonalds from our data, which is the largest firm and largest Media/outlet expenditures by far, theMedia/outlet coefficients rise modestly, but are little changed.22 Note that these regressions use the “more conservative” weighted between Tobit estimator, and do not includesectoral dummy variables, because we want to explain the differences across sectors within quantiles, though addingsectoral dummies does not change the nature of the results.23 Because adding all these variables jointly considerably lowers the sample size (data for different variables aremissing for different chains), we add these variables one at a time and then jointly. When we add them jointly, themuch smaller sample size (falling to 930) does lower the significance of all variables.
32
externalities. Consequently, we find, consistent with the existing literature, that franchisors
protect the brand by operating more of the units directly.24 Also, franchisors maintain more
company units when they require that franchisees have previous experience (Experience
Required) in the business. Again these are businesses where downstream operations are likely to
be more complicated and so the franchisor protects its tradename by maintaining direct control
over more units.
TABLE 6:PERCENT COMPANY OWNED REGRESSIONS WITH ADDITIONAL EXPLANATORY VARIABLES
TOBIT BETWEEN TOBIT
All SectorDummies
All SectorsDummies
TRAINING REQUIRED N=5693
.360(17.61)
χ2(33)=1117
.270(12.20)
χ2(55)=1431
.34(8.40)
χ2(33)=314
.265(5.83)
χ2(55)=393
NUMBER OF EMPLOYEES N=3547
.452(15.49)
χ2(25)=710
.417(10.98)
χ2(47)=925
.430(8.93)
χ2(33)=270
.428(6.68)
χ2(55)=351
CAPITAL REQUIRED N=7670
1.0(5.16)
χ2(33)=1024
.57(2.99)
χ2(55)=1431
.90(2.45)
χ2(33)=255
.68(1.79)
χ2(55)=378
EXPERIENCE REQUIRED N=4073
6.05(6.90)
χ2(27)=662
5.19(5.91)
χ2(49)=1020
8.26(4.77)
χ2(33)=206
7.08(4.01)
χ2(55)=285
NUMBER OF STATES N=5950
-.347(-13.96)
χ2(34)=993
.281(-10.49)
χ2(56)=1380
-.305(-6.23)
χ2(34)=280
-.234(-4.43)
χ2(56)=391
FINANCINGAVAILABLE N=7779
-2.92(-9.37)
χ2(33)=1085
-1.99(-6.51)
χ2(55)=1704
-3.41(-4.24)
χ2(33)=268
-2.33(-2.89)
χ2(55)=386
CONTRACT LENGTH N=3690
.293(4.37)
χ2(25)=549
.027(0.38)
χ2(47)=885
.201(2.14)
χ2(33)=214
-.011(-0.11)
χ2(55)=315
Each cell represents a separate regression; the coefficient (t-statistic) is displayed for the variable listed. Each regressionalso contains the variables listed in regressions of Table 4: Media/outlet in quartic form; advertising/outlet; years before;years before-squared; outlets; outlets-squared; Canadian dummy; year dummies; franchising entry-year dummies.
24 See Lafontaine and Slade (1997) for a review of the existing literature that shows that this empirical result, thatchains with larger units operate more of them directly, is robust across studies in the literature.
33
Interestingly, neither of these effects, of outlet size and of previous experience required,
on company ownership, is consistent with standard franchisee monitoring arguments. First,
larger outlets, whose management requires previous experience, likely means a more difficult job
for the manager, so that more franchising should occur in a standard agency model. Second,
when outlets are physically large, they tend to be less densely located, with the result that
monitoring each outlet is more costly, requiring again more incentive contracting. Our data,
however, clearly show that franchisors target more company ownership when each outlet is large
and when the manager’s job requires previous experience.
Our results on geographical dispersion—the Number of States with outlets—confirm that
firms operate less units directly when they are more geographically dispersed. This result is
consistent with the use of franchising to give incentives to outlet managers whose actions are
difficult to monitor. This finding has been very robust in the literature to date.
Finally, a few words about franchisor financing and contract length. On the former, the
negative effect of the provision of Financing Available on the “stable” proportion of company
units suggests that this financing relaxes franchisee capital constraints, allowing the franchisor to
franchise more. This is in contrast to the argument in the literature where franchisees are viewed
as a source of capital for their franchisors. However, as we are dealing with established
franchisors, it is not so surprising that they need not rely on franchisee capital for their expansion
needs. In fact, such franchisors must find franchising especially valuable if they are willing to
provide financing – after all, their capacity to finance franchisees suggests that they have the
resources necessary to operate corporately but simply choose not to. As for Contract Length,
from a control of the brand perspective, one might expect that long-term contracts are inflexible,
so that franchisors should use them less (own a greater proportion of units) if they need to protect
34
the brand. Our data show that indeed there is a greater tendency for firms using longer-term
contracts to rely on company ownership more. However, the effect of Contract Length becomes
insignificant when sector dummy variables are included in the regression. Thus, the effect of
Contract Length mostly occurs across sectors and is well captured by sector fixed effects.
Though not shown in Table 6, it is important to point out that our results concerning our
three main measures of tradename value are basically unaffected by the addition of the new
explanatory variables of Table 6. In all cases, years spent developing the business before
franchising and media expenditures per outlet continue to have strong positive effects on the
“stable” proportion of company units. The initial weakness of the Advertising effect means that
it does become insignificant in smaller samples for between Tobits.
4.4 TARGETING MANAGERIAL CONTROL : SECTORAL DIFFERENCES
Our regression results clearly show important sectoral differences in the reliance on
company ownership. Most likely, these reflect differences in production technology that affect
the incentive issues central to franchising. Here we explore these sectoral effects in more detail.
Figure 4 shows the distribution of the proportion of company units for aggregate
franchising sectors, with the sectors shown in decreasing order of the proportion of firms within
each that have less than 5 percent of company units.25 This figure shows remarkable variation in
the "target" proportion of company units across these aggregate sectors: the average proportion
of company ownership varies importantly across sectors. In a sizable number of sectors, we find
very little company ownership — see the results for travel, real estate, contractors, and printing
services, where the vast majority of the firms in the sector have less than 5 percent company
25 The data used in this section is for all franchising firms, not just those with more than seven years of experience asin the previous section.
35
units. About 25 percent of our data fall in sectors with mean company ownership less than ten
percent. In contrast, at the other end of the distribution lie all eating establishments—fast food
and full service restaurants—as these all have very high levels of company ownership.
FIGURE 4: THE DISTRIBUTION OF THE PROPORTION OF COMPANY UNITS ACROSS MAIN
AGGREGATE SECTORS
Health & Fitness
Eating Places - Full Svc
Retail - Home Furnishings
Hotels & Motels
Eating Places - Limited Svc
Retail - Food
Retail - Clothing
Cosmetic
Personal Services
Retail - Other
Business Services
Repair
Automotive
Rental
Retail - Used Products
Retail - Building Materials
Education
Maintenance
Recreation
Business Supplies
Contractors
Real Estate
Travel
0.00 20.00 40.00 60.00 80.00 100.00
pcentc<5.0
5.0<=pcentc<15.0
15.0<=pcentc<50.0
50.0<=pcentc
36
However, Figure 4 also shows significant variation within sectors, as there are firms in
each category of percent company ownership in all of the sectors. To determine whether the
within-sector variation is due to heterogeneous sector definitions, which would allow for very
different types of businesses and production technology to be grouped in the aggregate sectors,
we classified all the franchisors in our data into a set of detailed, very homogeneous sectors. For
example, in the automotive services sector, we separated out firms involved in “Detailing,
Window Tinting, Rustproofing” from “Mufflers Shops” from “Oil Change & Lube: Shops”
from “Transmission repair” and from “Auto Parts Retailing.” In total, we created, in this
process, 271 separate detailed sectors for business-format franchising. However, because these
sectors are so detailed, in many cases we have very few firms per sector. Restricting ourselves to
sectors that have 15 or more firms per sector in our data set, the number of sectors falls to 127,
and if we restrict it to those with 20 or more firms, the number of detailed sectors falls to 104.
Figure 5 displays distributions for the proportion of company units for the largest 25 of
those narrowly defined sectors, where the order is again based on the number of firms in the 0 to
5% category. This figure shows that the variation across such sectors is possibly larger than it
was across aggregate sectors. However, the within-sector differences remain very large as well.
To determine the extent of sectoral differences in PCENTC, we created a new data set of
sample size 271 containing sector-specific means for all our variables. We then regressed sector
means of PCENTC on sector means of all the X variables introduced above. We found that three
variables were significant in explaining sectoral differences: media, the number of states, and
whether experience was required.26
26 For the sample size of 104 sectors, the coefficients (t-statistics) are: .0037 (2.76)*media/outlet - .004(-3.08)*states+ .083(2.18) *experience +.00004(1.69) * outlets. Regression results are more significant if more sectors are added.
37
FIGURE 5: THE DISTRIBUTION OF THE PROPORTION OF COMPANY UNITS
ACROSS THE LARGEST 25 NARROWLY DEFINED SECTORS
Temporary Help - Others NEC
Fast Food - Hamburgers
Auto - Oil Change & Lube: Shops
Fast Food - Chicken
Family Restaurants
Fast Food - Others NEC
Italian Restaurants
Food - Retail: Convenience Stores
Fast Food - Sandwiches, Soups & Salads
Fast Food - Pizza
Hotels & Motels
Fast Food - Mexican
Fast Food - Donuts
Auto - Mufflers Shops
Cosmetic - Hair Care
Fast Food - Submarine Sandwiches
Employment Agencies - Others NEC
Carpet, Upholstery & Drapery Cleaning
Rental - Car
Food - Snacks: Ice Cream
Accounting, Bookkeeping & Tax Services
Maintenance: Janitorial
Auto - Repair: Transmission
Brokers & Real Estate Agents
Printing Services
0.00 20.00 40.00 60.00 80.00 100.00
pcentc<5.0
5.0<=pcentc<15.0
15.0<=pcentc<50.0
50.0<=pcentc
We now turn back to the regressions in Table 4 and introduce detailed sectoral controls in
our basic PCENTC regressions. We limit our data set to those firms that belong to one of the
104 detailed sectors with 20 or more firms. As a result, our data set falls from 1563 to 1180
These also contain the variables years-before-franchising and its squared term, advertising/outlet, and year dummies.
38
firms. When we add the 23 aggregate sector dummy variables to the percent company owned
regression for this data set, the R-squared rises from .17 for the basic regression to .22. If we
instead introduce fixed effects for the 104 detailed sectors the R-squared rises to .35 (or .27 for
the adjusted R-squared).27 As in Table 4, the media effect falls very slightly when the main
sector dummies are added, and falls slightly more when the detailed sector dummies are added,
but in all cases the media effect remains very highly significant.
In sum, we find that the detailed sectoral dummies do add to the explanatory power of the
PCENTC regression, but they do not really dampen the media effect, which appears to apply
both within narrowly defined sectors and across them. These results are consistent with the
sectoral differences shown in Figures 4 and 5—sectoral effects are significant in our regressions,
but most of the explained variation in our data occurs within sectors for all explanatory variables
other than media, states, and contract length.
4.5 TARGETING MANAGERIAL CONTROL : WITHIN -SECTOR EVIDENCE
Having shown that within sector variation is an important factor in our data, we now present
examples of the targets that are set by firms in a few of the largest franchising sectors, including
fast food and auto rental. We are in the very unique position of being able to see and publish the
company names for our large data set of more than 1500 firms that have extensive franchising
experience—in most other firm-level research with large samples the company names are
confidential. The examples below are enlightening and particularly useful in reinforcing the
statistical results presented above.
27 Note that if we use the full sample of 1563 and 271 sectoral dummies the R-squared also rises only to .35 (.32adjusted R-squared). These regressions are between OLS regressions so we can compare the R-squared after addingsectoral dummies, but as is evident in Table 4, there is little difference between the tobits and the OLS models.
39
TABLE 7: COMPANY LEVEL DATA
Company %Co-own
Outlets Media/outlet
# Obs. Company %Co-own
Outlets Media/outlet
# Obs.
Car Rentals Chicken
Hertz Rent-A-Car 66.1 1601 9.07 4 Church's Chicken 67.3 1310 4.02 17
National Car Rental 39.5 951 9.78 4 Grandy's 52.1 164 15.63 10
Budget Rent A Car 8.0 2937 3.99 17 Bojangles of Amer. 48.6 244 7.42 9
Thrifty Rent-A-Car 2.6 699 5.22 17 El Pollo Loco 38.8 245 35.71 1
Dollar Rent A Car 1.8 1501 2.04 15 KFC 26.1 7976 12.18 16
Rent-A-Wreck 1.4 314 0.25 15 Pioneer Take Out 24.1 288 5.35 9
Ugly Duckling Rent-A-Car System
0.5 363 0.50 9 Popeye's Chicken &Biscuits
16.4 665 8.57 17
Holiday-PaylessRent-A-Car System
0.3 182 0.00 7 Lee's Famous RecipeCountry Chicken
14.6 262 3.47 13
American Intl RentA Car
0.1 1294 0.09 11 Famous Recipe FriedChicken
5.8 225 1.17 2
Econo Car 0.0 161 0.00 2 Golden Skillet FriedChicken
2.6 159 0.96 6
Hamburger Pizza
Carl's Jr. Restaurant 84.7 472 13.02 8 Pizza Hut 49.2 4549 11.77 7
Krystal Restaurants 81.0 308 17.87 6 Domino's Pizza 29.6 2919 10.90 13
Jack in the Box 80.3 864 21.10 9 Pizza Inn 27.8 610 2.40 15
What A Burger 57.2 344 5.50 7 Little Caesars Pizza 27.7 1777 14.91 14
Checkers Drive-InRestaurants
56.6 169 22.71 4 Dino's Pizza 13.3 167 0.39 5
Burger Chef 37.0 710 7.79 3 Shakey's PizzaRestaurant
6.9 391 5.50 13
Wendy's OldFashioned Hamb.
31.8 3084 23.15 13 Round Table Pizza 1.3 472 5.36 18
Burger Queen 31.2 192 3.56 1 Hungry Howie'sPizza & Subs
0.0 223 1.98 8
McDonald's 26.5 11067 33.22 18
Rally's Hamburgers 25.9 282 15.36 6
Burger King 14.1 5366 28.79 14
Sonic Drive-InRestaurant
9.8 1155 5.30 16
Note: Excludes franchisor with no media data, those with less than 150 units, and Canadian franchisors..
40
In Table 7 we display the company-specific time-series mean values for the key variables
(PCENTC, outlet numbers, and media/outlet) in four detailed sectors—namely car rental and
then three types of fast-food, chicken, hamburger and pizza—that have high levels of company
ownership. We sort the companies by decreasing proportions of company units, and restrict the
sample to those firms with more than 150 outlets and 8 years of franchising experience (to limit
the size of the table and display only firms committed to franchising).
The most striking evidence comes from auto rental companies. The two companies with the
largest per outlet media values, Hertz, and National, target extraordinarily high levels of
company ownership (66% and 40% respectively). In contrast, a large number of firms target
company ownership shares from 0 to 3% (see American International, Dollar, Holiday, Rent-a-
Wreck, Thrifty, U-Save, and Ugly Duckling). These firms are sizable franchisors, but are
investing less in media, thus for those with cross-state coverage, PCENTC is very low.
Moreover, the firms that most people perceive as advertising high-quality service (Hertz and
National) have much higher levels of company ownership than those firms that advertise low-
cost service (such as Budget, Thrifty and Dollar). Similarly, in the fast-food chicken sector, firms
that target high levels of company ownership (above 15%) clearly advertise more, per outlet, on
average than those with low levels of company ownership. The same is true in the pizza
industry, where the proportion of company units for the three best-known and largest chains,
Pizza Hut, Domino’s, and Little Ceasar’s, are clearly well above those of most chains in the
segment. These firms also tend to advertise more than most, on a per outlet basis. Yet it is also
true that those smaller pizza chains that also spend a lot on advertising per outlet tend to have
high proportions of company units. The evidence is least clear in the hamburgers sector. But
41
even here, we see that the established firms, McDonald’s and Wendy’s, target company
ownership levels above 25 percent.28 With the exception of Burger King, it is also clear that
firms that have high media expenditures per outlet tend to have high proportions of company
units.
As we discussed managerial targeting in this section, we presented only averages of the
proportions of company ownership over time – space limitations prevent us from displaying
time-series company-level data. Of course, our earlier results supported the lack of time-series
variation in the proportion of company units in our overall sample, allowing us to concentrate on
time-series means. Still, it is worth pointing out again that the companies with high targets of
PCENTC that we mention here have especially stable levels of this over time.
In sum, the company examples deepen our interpretation of our Media variable effects on
company targeting. We see that companies that target high levels of company ownership due to
high Media effects are targeting high-quality tradename values.
5. CONCLUSION
Using an extensive longitudinal data set on franchising firms, we show that established
franchisors manage their portfolio of company and franchised units to maintain a particular
target level of corporate control and ownership. That is, stable franchisors, having eight or more
years of experience and at least 15 outlets, maintain a constant percent of company outlets as the
franchisor grows or gains experience—company ownership of outlets does not rise or fall with
experience or learning, or with success or failure. On average, established franchisors maintain
28 Note that two companies, Carl’s Junior and Jack in the Box, have high levels of company ownership because theystarted franchising much later than they started their business (both started franchising in early 1980’s and beganbusiness in 1941 and 1951, respectively). During our sample period, as they began franchising, their company
42
about 15 percent of their outlets as company owned and the other 85 percent are owned by their
franchisees. However, this average rate of company ownership varies considerably across
sectors, with average rates as high as 26 percent for restaurants and 16 percent for fast food, and
as low as 2 percent for travel services. The targeted rate also varies considerably across firms:
firm-specific fixed effects explain 90 percent of the variance of company ownership rates in our
longitudinal database.
Past empirical work had not identified this practice of targeting a fixed rate of company
ownership, and had instead concluded that company ownership declines with experience.
Company ownership does decline for the first seven years of experience. Since franchisors start
with 100 percent company owned outlets, this proportion can only go down as they add
franchised outlets. We find that by the end of seven years, on average, companies have achieved
their target rate of corporate ownership and thereafter this rate becomes very stable. However, as
mentioned earlier, this conclusion that franchisors choose a particular level of company
ownership does not preclude franchisors from choosing which units should be franchised and
which company operated.
Given strong evidence that firms target specific but different rates of company
ownership, what factors determine firms’ optimal targeted rates? We show that brandname
value is a primary determinant: franchisors with high brandname value target high rates of
company ownership. High rates of company ownership are needed in chains with more valuable
brands because each franchisee has an incentive to free-ride on brandname value or undermine
brandname value. As each franchisee maximizes outlet profits, he or she receives only a small
portion of the return to investing in brandname value, and the remaining return accrues to other
ownership falls from 99% to about 60% for both, at which point our time-series data ends.
43
franchisees or to the franchisor. Thus, when brandname value is high, franchisors must exert
more managerial control through their own outlets, by raising rates of company ownership of
outlets. Moreover, higher brandname values imply a central role for franchisor effort in
generating revenues and furthering the brand through advertising, new product development, and
related activities. High rates of company ownership give franchisors a large stake in the business
and thus higher-powered incentives to undertake these activities. Hence, both in terms of
controlling franchisee free-riding, and in terms of franchisor incentives, one expects higher rates
of company ownership in franchised systems with more valuable brands. When brandname
value is measured as expenditures on media or advertising, we find that these measures are
highly significant determinants of company ownership rates, explaining 11 percent of the
variance in company ownership rates across firms.
44
REFERENCES
Anderson, E.E. (1984), “The Growth and Performance of Franchise Systems: Company versusFranchisee Ownership,” Journal of Economics and Business, 36, 421-431.
Bai, Chong-en and Zhigang Tao, (2000), “Contract Mix in Franchising,” Journal of Economics& Management Strategy, vol. 9, n.1, p. 85-113.
Bercovitz, J.E.L. (1998b), “Franchising vs. Company Ownership,” mimeo, Fuqua School ofBusiness, Duke University.
Bhattacharyya, S. and F. Lafontaine (1995), “Double-Sided Moral Hazard and the Nature ofShare Contracts,” RAND Journal of Economics, 26, 761-781.
Blair, R. D. and D. L. Kaserman, (1982), “Optimal Franchising,” Southern Economic Journal,494-504.
Blair, R.D. and D. L. Kaserman, (1994), “A Note on Incentive Incompatibility underFranchising,” Review of Industrial Organization, 9, 323-330.
Bond, R. (1997), Bond’s Franchising Guide, SourceBook Publications.Bradach, J.L. (1997), “Using the Plural Form in the Management of Restaurant Chains,”
Administrative Science Quarterly, 42, 276-303.Bradach, J. L. and R. Eccles (1989), “Price, Authority and Trust,” Annual Review of Sociology,
15, 97-118.Brickley, J.A., (1999), “Incentive Conflicts and Contracting: Evidence from Franchising,”
Journal of Law and Economics, 42, 745-774.Brickley, J. and F. Dark (1987), “The Choice of Organizational Form: The Case of Franchising,”
Journal of Financial Economics, 18, 401-420.Brickley, J.A., F.H. Dark, and M.S. Weisbach (1991a), “An Agency Perspective on
Franchising,” Financial Management, 20, 27-35.Brickley, J.A., F.H. Dark, and M.S. Weisbach, (1991b), “The Economic Effects of Franchise
Termination Laws,” Journal of Law and Economics, 34, 101-132.Carmichael, H.L. (1983), “The Agent-Agents Problem: Payment by Relative Output,” Journal of
Labor Economics, 1, 50-65.Carney, M. and E. Gedajlovic (1991), “Vertical Integration in Franchise Systems: Agency
Theory and Resource Explanations,” Strategic Management Journal, 12, 607-629.Caves, R.E. and W.F. Murphy (1976), “Franchising: Firms, Markets, and Intangible Assets,”
Southern Economic Journal, 42, 572-586.Dant, R.P., P.J. Kaufmann and A.K. Paswan (1992), “Ownership Redirection in Franchised
Channels,” Journal of Public Policy and Marketing, 11, 33-44.Dnes, A. W. (1992a), Franchising: A Case-study Approach, Ashgate Publishing Ltd., Aldershot,
England.Dnes, A. W. (1992b), “Unfair' Contractual Practices and Hostages in Franchise Contracts,”
Journal of Institutional and Theoretical Economics, 148, 484-504.Dnes, A.W. (1993), “A Case Study Analysis of Franchise Contracts”, Journal of Legal Studies,
22, 367-393.Dnes, A. W. (1996), “The Economic Analysis of Franchise Contracts,” Journal of Institutional
and Theoretical Economics, 152, 297-324.Elango, B. and V. H. Fried (1997), “Franchising Research: A Literature Review and Synthesis,”
Journal of Small Business Management, 35, 68-82.
45
Gallini, Nancy T., and Nancy A. Lutz. (1992), “Dual Distribution and Royalty Fees inFranchising,” Journal of Law, Economics, & Organization, 8, 471-501.
Hunt, S.D., (1973), “The Trend Toward Company-Operated Units in Franchise Chains,” Journalof Retailing, Summer, 3-12.
Kaufmann, P.J. and Lafontaine, F. (1994), “Costs of Control: The Source of Economic Rents forMcDonald's Franchisees,” Journal of Law and Economics, 37, 417-543.
Kehoe, M. R. (1996), “Franchising, agency problems, and the cost of capital,” AppliedEconomics, 28, 1485-1493.
Klein, B. (1995), “The Economics of Franchise Contracts,” Journal of Corporate Finance, 2, 9-37.
Lafontaine, F. (1992a), “Agency Theory and Franchising: Some Empirical Results,” RANDJournal of Economics, 23, 263-283.
Lafontaine, F. (1992b), “How and Why do Franchisors do What They do: A Survey Report”, inP.J. Kaufmann (Ed.), Franchising: Passport for Growth & World of Opportunity, SixthAnnual Proceedings of the Society of Franchising, University of St-Thomas.
Lafontaine, F. and Bhattacharyya, S., (1995), “The Role of Risk in Franchising”, Journal ofCorporate Finance, 2, 39-74.
Lafontaine, F. and P.J. Kaufmann (1994), “The Evolution of Ownership Patterns in FranchiseSystems,” Journal of Retailing, 70, 97-113.
Lafontaine, F. and K.L. Shaw, (1998), “Franchising Growth and Franchisor Entry and Exit in theU.S. Market: Myth and Reality,” Journal of Business Venturing, 13,95-112.
Lafontaine, F. and K.L. Shaw, (1999) “The Dynamics of Franchise Contracting: Evidence fromPanel Data”, Journal of Political Economy, 107, 1041-1080.
Lafontaine, F. and Slade, M.E. (1996), “Retail Contracting and Costly Monitoring: Theory andEvidence”, European Economic Review Papers and Proceedings, 40, 923-932.
Lafontaine, F. and Slade, M.E. (1997), “Retail Contracting: Theory and Practice,” Journal ofIndustrial Economics, 45, 1-25.
Lafontaine F. and Slade, M.E., (2001), “Incentive Contracting and the Franchise Decision,” forin Advances in Business Applications of Game Theory, K. Chatterjee and W. Samuelson(eds.) Kluwer Academic Press.
Lal, R. (1990), “Improving Channel Coordination through Franchising,” Marketing Science, 9,299-318.
Lewin, S. (1998), "Innovation and Authority in Franchise Systems: An Exploration of the PluralForm,” mimeo, Iowa State University.
Lillis, C.M., Narayana, C.L. and J.L. Gilman, (1976), “Competitive Advantage Variation Overthe Life Cycle of a Franchise,” Journal of Marketing, 40, 77-80.
Love, J.F., (1986), McDonald's: Behind the Arches, Bantam Books, Inc. New York.Lutz, N.A. (1995), “Ownership Rights and Incentives in Franchising,” Journal of Corporate
Finance, 2, 103-130.Luxenberg, S., (1985), Roadside Empires: How the Chains Franchised America, Penguin Books.Lyons, B.R. (1996), “Empirical Relevance of Efficient Contract Theory: Inter-Firm Contracts,”
Oxford Review of Economic Policy,” 12, 27-52.Maness, R. (1996), “Incomplete Contracts and the Choice Between Vertical Integration and
Franchising,” Journal of Economic Behavior and Organization, 31, 101-115.Martin, R.E. (1988), “Franchising and Risk Management,” American Economic Review, 78, 954-
968.
46
Mathewson, F. and R. Winter (1985), “The Economics of Franchise Contracts,” Journal of Lawand Economics, 28, 503–526.
Minkler, A. (1990), “An Empirical Analysis of a Firm's Decision to Franchise,” EconomicsLetters, 34, 77-82.
Minkler, A. and T. A. Park (1994), “Asset Specificity and Vertical Integration,” Review ofIndustrial Organization, 9, 409-423.
Muris, T.J., D.T. Scheffman, and P.T. Spiller (1992), “Strategy and Transaction Costs: TheOrganization of Distribution in the Carbonated Soft Drink Industry,” Journal ofEconomics & Management Strategy, 1, 83-128.
Noren, D.L., (1990), “The Economics of the Golden Arches: A Case Study of the McDonald'sSystem,” The American Economist, 34, 60-64.
Norton, S.W. (1988), “An Empirical Look at Franchising as an Organizational Form” Journal ofBusiness, 61: 197-217.
Oxenfeldt, A.R and A.O. Kelly, (1969), "Will Successful Franchise Systems Ultimately BecomeWholly-Owned Chains?", Journal of Retailing, 44, 69-87.
Ozanne, U.B. and S.D. Hunt, (1971), The Economic Effect of Franchising, U.S. Senate, SelectCommittee on Small Business, Washington, Gov. Printing Office.
Rubin, P. (1978), “The Theory of the Firm and the Structure of the Franchise Contract,” Journalof Law and Economics, 21, 223-233.
Sasser, W.E. and S.H. Pettway, (1974), “Case of Big Mac's Pay Plans,” Harvard BusinessReview, 52, 30-46 and 156-158.
Schmidt, Torsten (1994), “An Analysis of Intrabrand Competition in the Franchise Industry”Review of Industrial Organization, 9, 293-310.
Scott, F.A. (1995), “Franchising vs. Company Ownership as a Decision Variable of the Firm,”Review of Industrial Organization, 10, 69-81.
Shelton, J. (1967), “Allocative Efficiency vs. ‘X-Efficiency’: Comment,” American EconomicReview, 57, 1252-1258.
Shepard, A. (1993), “Contractual Form, Retail Price, and Asset Characteristics,” Rand Journal ofEconomics, 24, 58-77.
Slade, M.E. (1996), “Multitask Agency and Contract Choice: An Empirical Assessment,”International Economic Review, 37, 465-486.
Thompson, R.S. (1994), “The Franchise Life Cycle and the Penrose Effect,” Journal ofEconomic Behavior and Organization, 24, 207-218.
U.S. Department of Commerce, (1998)Franchising in the Economy, 1986-1988, prepared byAndrew Kostecka.
47
Data Appendix
Our franchisor-level data come from two main sources. For the period from 1980 to 1992
inclusively, the main source of data is the Entrepreneur magazine’s “Annual Franchise 500”
surveys. In each case, the data for a given year are obtained from the following year survey as
these are all published early in the year. In its 1994 survey, which provides data for 1993, the
magazine covered fewer firms than usual, and in the 1995 survey onward, it stopped reporting
advertising fee data. For those reasons, starting with the 1993 data, we use the Source Book of
Franchise Opportunities, now called Bond’s Franchise Guide, as our main source of data. This
has the added advantage that there is more detailed information on each franchisor in our data
from that point onward.
In Table A1, we describe the sample according to the number of years over which we
observe each firm in our sample. The change in the composition of our sample of franchisors
over time is due first and foremost to entry and exit from franchising. The data reveal that about
15 percent of the franchisors covered in the survey each year have just started franchising. Also,
there is considerable exit from franchising: about 140 franchising firms stop operations each year
according to the U.S. Commerce Department (1988). The Franchise Annual and the Bond’s
Franchise Guide report even higher exit rates than that. However, our panel is unbalanced also
because we have been fairly conservative in matching firms across years. If we could not find an
exact or very convincing match in terms of firm name (and address) over time, we kept the firms
separate. Finally, of course, the data are unbalanced also because firms that answer these
surveys one year may decide not to respond the next year. Firms are more likely to respond to
the surveys when they wish to grow and thus find value in the visibility that these listings
provide. Thus, our data are biased in favor of new or expanding firms.
48
TABLE A1: SAMPLE CHARACTERISTICS
Observed Number of Firms Number ofObservations
StartedFranchising in
(mean)
Started inBusiness in
(mean)
Number of Firms(ConsecutiveObservations)
1 time 1461 1461 1984.1 1977.6 1461
2 times 944 1888 1984.3 1978.0 863
3 times 578 1734 1982.8 1976.8 478
4 times 475 1900 1981.9 1974.4 362
5 times 293 1465 1981.7 1975.1 193
6 times 216 1296 1979.6 1972.6 144
7 times 176 1232 1979.4 1972.6 96
8 times 136 1088 1978.1 1970.3 74
9 times 92 828 1978.7 1969.3 41
10 times 76 760 1979.2 1970.2 37
11 times 82 902 1977.8 1973.1 41
12 times 59 708 1976.5 1969.4 30
13 times 56 728 1972.8 1965.1 21
14 times 38 532 1973.0 1965.9 16
15 times 41 615 1974.3 1968.4 14
16 times 40 640 1969.3 1962.6 19
17 times 38 646 1969.9 1963.8 12
18 times 41 738 1965.7 1961.8 41
Total 4842 19162 1982.0 1975.4 3943
Top Related